The exponentially weighted moving average chart of the squared deviation (EWMAS) is often applied for monitoring changes such as step shifts and linear drifts in process variation when no subgrouping is available. This paper analyzes the performance of the EWMAS chart under drifts in process variation. A fast and accurate algorithm based on the piecewise collocation method is presented for computing both the zero-state and steadystate average run lengths of the EWMAS chart. It is shown that the proposed method can provide accurate approximation results in both zero-state and steady-state cases. Some optimal design tables are also provided to facilitate the design of EWMAS charts in practice.